NextGen PDF — High-Fidelity Office → PDF Fix

This patch fixes the core Office-to-PDF problem in the NextGen PDF backend.

What was wrong

The previous /api/word-to-pdf, /api/excel-to-pdf, and /api/pptx-to-pdf implementations extracted text with Python libraries and manually drew that text into a new PDF. That discards much of the original document layout.

What this patch changes

Office documents are now rendered by LibreOffice:

DOCX/DOC → LibreOffice → PDF XLSX/XLS → LibreOffice → PDF PPTX/PPT → LibreOffice → PDF

This is a document renderer, not a text reconstruction pipeline. It therefore preserves the original document's layout much more faithfully, including images, fonts available on the server, paragraph formatting, tables, margins, headers/footers, page breaks and spacing.

Files

• backend/office_converter.py — reusable conversion engine
• backend/Dockerfile — installs LibreOffice + fonts on Render
• backend/requirements.txt — backend dependencies
• render.yaml — switches the Render service to Docker
• apply_conversion_fix.py — safely patches the existing backend/main.py

Apply to your current project

Copy these files into the root of your current NextGen PDF project, then run:

python apply_conversion_fix.py

The script creates:

backend/main.py.before-office-fix

before changing backend/main.py.

Then verify:

python -m py_compile backend/main.py backend/office_converter.py

Local requirement

For local Word/Excel/PowerPoint → PDF testing, install LibreOffice on the machine and make sure libreoffice or soffice is on PATH.

Render deployment

The included render.yaml uses the Dockerfile so Render installs LibreOffice. This is required; simply adding a Python package is not enough.

After deployment, test with the original Word assignment and compare the PDF page-by-page with Word.

Important

Do not use the old text-reconstruction implementations for Office → PDF again. All three Office conversion endpoints should use convert_office_to_pdf().